vde-2/slirpvde/tcp_timer.c

   1 /*
   2  * Copyright (c) 1982, 1986, 1988, 1990, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  *
  33  *      @(#)tcp_timer.c 8.1 (Berkeley) 6/10/93
  34  * tcp_timer.c,v 1.2 1994/08/02 07:49:10 davidg Exp
  35  */
  36
  37 #include <config.h>
  38 #include <slirp.h>
  39
  40 #define max(x,y) ((x) > (y) ? (x) : (y))
  41 #define min(x,y) ((x) < (y) ? (x) : (y))
  42
  43 int     tcp_keepidle = TCPTV_KEEP_IDLE;
  44 int     tcp_keepintvl = TCPTV_KEEPINTVL;
  45 int     tcp_maxidle;
  46 int     so_options = DO_KEEPALIVE;
  47
  48 struct   tcpstat tcpstat;        /* tcp statistics */
  49 u_int32_t        tcp_now;                /* for RFC 1323 timestamps */
  50
  51 /*
  52  * Fast timeout routine for processing delayed acks
  53  */
  54 void
  55 tcp_fasttimo()
  56 {
  57         register struct socket *so;
  58         register struct tcpcb *tp;
  59
  60         DEBUG_CALL("tcp_fasttimo");
  61
  62         so = tcb.so_next;
  63         if (so)
  64         for (; so != &tcb; so = so->so_next)
  65                 if ((tp = (struct tcpcb *)so->so_tcpcb) &&
  66                     (tp->t_flags & TF_DELACK)) {
  67                         tp->t_flags &= ~TF_DELACK;
  68                         tp->t_flags |= TF_ACKNOW;
  69                         tcpstat.tcps_delack++;
  70                         (void) tcp_output(tp);
  71                 }
  72 }
  73
  74 /*
  75  * Tcp protocol timeout routine called every 500 ms.
  76  * Updates the timers in all active tcb's and
  77  * causes finite state machine actions if timers expire.
  78  */
  79 void
  80 tcp_slowtimo()
  81 {
  82         register struct socket *ip, *ipnxt;
  83         register struct tcpcb *tp;
  84         register int i;
  85
  86         DEBUG_CALL("tcp_slowtimo");
  87
  88         tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
  89         /*
  90          * Search through tcb's and update active timers.
  91          */
  92         ip = tcb.so_next;
  93         if (ip == 0)
  94            return;
  95         for (; ip != &tcb; ip = ipnxt) {
  96                 ipnxt = ip->so_next;
  97                 tp = sototcpcb(ip);
  98                 if (tp == 0)
  99                         continue;
 100                 for (i = 0; i < TCPT_NTIMERS; i++) {
 101                         if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
 102                                 tcp_timers(tp,i);
 103                                 if (ipnxt->so_prev != ip)
 104                                         goto tpgone;
 105                         }
 106                 }
 107                 tp->t_idle++;
 108                 if (tp->t_rtt)
 109                    tp->t_rtt++;
 110 tpgone:
 111                 ;
 112         }
 113         tcp_iss += TCP_ISSINCR/PR_SLOWHZ;               /* increment iss */
 114 #ifdef TCP_COMPAT_42
 115         if ((int)tcp_iss < 0)
 116                 tcp_iss = 0;                            /* XXX */
 117 #endif
 118         tcp_now++;                                      /* for timestamps */
 119 }
 120
 121 /*
 122  * Cancel all timers for TCP tp.
 123  */
 124 void
 125 tcp_canceltimers(tp)
 126         struct tcpcb *tp;
 127 {
 128         register int i;
 129
 130         for (i = 0; i < TCPT_NTIMERS; i++)
 131                 tp->t_timer[i] = 0;
 132 }
 133
 134 int     tcp_backoff[TCP_MAXRXTSHIFT + 1] =
 135    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
 136
 137 /*
 138  * TCP timer processing.
 139  */
 140 struct tcpcb *
 141 tcp_timers(tp, timer)
 142         register struct tcpcb *tp;
 143         int timer;
 144 {
 145         register int rexmt;
 146
 147         DEBUG_CALL("tcp_timers");
 148
 149         switch (timer) {
 150
 151         /*
 152          * 2 MSL timeout in shutdown went off.  If we're closed but
 153          * still waiting for peer to close and connection has been idle
 154          * too long, or if 2MSL time is up from TIME_WAIT, delete connection
 155          * control block.  Otherwise, check again in a bit.
 156          */
 157         case TCPT_2MSL:
 158                 if (tp->t_state != TCPS_TIME_WAIT &&
 159                     tp->t_idle <= tcp_maxidle)
 160                         tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
 161                 else
 162                         tp = tcp_close(tp);
 163                 break;
 164
 165         /*
 166          * Retransmission timer went off.  Message has not
 167          * been acked within retransmit interval.  Back off
 168          * to a longer retransmit interval and retransmit one segment.
 169          */
 170         case TCPT_REXMT:
 171
 172                 /*
 173                  * XXXXX If a packet has timed out, then remove all the queued
 174                  * packets for that session.
 175                  */
 176
 177                 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
 178                         /*
 179                          * This is a hack to suit our terminal server here at the uni of canberra
 180                          * since they have trouble with zeroes... It usually lets them through
 181                          * unharmed, but under some conditions, it'll eat the zeros.  If we
 182                          * keep retransmitting it, it'll keep eating the zeroes, so we keep
 183                          * retransmitting, and eventually the connection dies...
 184                          * (this only happens on incoming data)
 185                          *
 186                          * So, if we were gonna drop the connection from too many retransmits,
 187                          * don't... instead halve the t_maxseg, which might break up the NULLs and
 188                          * let them through
 189                          *
 190                          * *sigh*
 191                          */
 192
 193                         tp->t_maxseg >>= 1;
 194                         if (tp->t_maxseg < 32) {
 195                                 /*
 196                                  * We tried our best, now the connection must die!
 197                                  */
 198                                 tp->t_rxtshift = TCP_MAXRXTSHIFT;
 199                                 tcpstat.tcps_timeoutdrop++;
 200                                 tp = tcp_drop(tp, tp->t_softerror);
 201                                 /* tp->t_softerror : ETIMEDOUT); */ /* XXX */
 202                                 return (tp); /* XXX */
 203                         }
 204
 205                         /*
 206                          * Set rxtshift to 6, which is still at the maximum
 207                          * backoff time
 208                          */
 209                         tp->t_rxtshift = 6;
 210                 }
 211                 tcpstat.tcps_rexmttimeo++;
 212                 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
 213                 TCPT_RANGESET(tp->t_rxtcur, rexmt,
 214                     (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
 215                 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
 216                 /*
 217                  * If losing, let the lower level know and try for
 218                  * a better route.  Also, if we backed off this far,
 219                  * our srtt estimate is probably bogus.  Clobber it
 220                  * so we'll take the next rtt measurement as our srtt;
 221                  * move the current srtt into rttvar to keep the current
 222                  * retransmit times until then.
 223                  */
 224                 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
 225 /*                      in_losing(tp->t_inpcb); */
 226                         tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
 227                         tp->t_srtt = 0;
 228                 }
 229                 tp->snd_nxt = tp->snd_una;
 230                 /*
 231                  * If timing a segment in this window, stop the timer.
 232                  */
 233                 tp->t_rtt = 0;
 234                 /*
 235                  * Close the congestion window down to one segment
 236                  * (we'll open it by one segment for each ack we get).
 237                  * Since we probably have a window's worth of unacked
 238                  * data accumulated, this "slow start" keeps us from
 239                  * dumping all that data as back-to-back packets (which
 240                  * might overwhelm an intermediate gateway).
 241                  *
 242                  * There are two phases to the opening: Initially we
 243                  * open by one mss on each ack.  This makes the window
 244                  * size increase exponentially with time.  If the
 245                  * window is larger than the path can handle, this
 246                  * exponential growth results in dropped packet(s)
 247                  * almost immediately.  To get more time between
 248                  * drops but still "push" the network to take advantage
 249                  * of improving conditions, we switch from exponential
 250                  * to linear window opening at some threshold size.
 251                  * For a threshold, we use half the current window
 252                  * size, truncated to a multiple of the mss.
 253                  *
 254                  * (the minimum cwnd that will give us exponential
 255                  * growth is 2 mss.  We don't allow the threshold
 256                  * to go below this.)
 257                  */
 258                 {
 259                 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
 260                 if (win < 2)
 261                         win = 2;
 262                 tp->snd_cwnd = tp->t_maxseg;
 263                 tp->snd_ssthresh = win * tp->t_maxseg;
 264                 tp->t_dupacks = 0;
 265                 }
 266                 (void) tcp_output(tp);
 267                 break;
 268
 269         /*
 270          * Persistence timer into zero window.
 271          * Force a byte to be output, if possible.
 272          */
 273         case TCPT_PERSIST:
 274                 tcpstat.tcps_persisttimeo++;
 275                 tcp_setpersist(tp);
 276                 tp->t_force = 1;
 277                 (void) tcp_output(tp);
 278                 tp->t_force = 0;
 279                 break;
 280
 281         /*
 282          * Keep-alive timer went off; send something
 283          * or drop connection if idle for too long.
 284          */
 285         case TCPT_KEEP:
 286                 tcpstat.tcps_keeptimeo++;
 287                 if (tp->t_state < TCPS_ESTABLISHED)
 288                         goto dropit;
 289
 290 /*              if (tp->t_socket->so_options & SO_KEEPALIVE && */
 291                 if ((so_options) && tp->t_state <= TCPS_CLOSE_WAIT) {
 292                         if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
 293                                 goto dropit;
 294                         /*
 295                          * Send a packet designed to force a response
 296                          * if the peer is up and reachable:
 297                          * either an ACK if the connection is still alive,
 298                          * or an RST if the peer has closed the connection
 299                          * due to timeout or reboot.
 300                          * Using sequence number tp->snd_una-1
 301                          * causes the transmitted zero-length segment
 302                          * to lie outside the receive window;
 303                          * by the protocol spec, this requires the
 304                          * correspondent TCP to respond.
 305                          */
 306                         tcpstat.tcps_keepprobe++;
 307 #ifdef TCP_COMPAT_42
 308                         /*
 309                          * The keepalive packet must have nonzero length
 310                          * to get a 4.2 host to respond.
 311                          */
 312                         tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
 313                             tp->rcv_nxt - 1, tp->snd_una - 1, 0);
 314 #else
 315                         tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
 316                             tp->rcv_nxt, tp->snd_una - 1, 0);
 317 #endif
 318                         tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
 319                 } else
 320                         tp->t_timer[TCPT_KEEP] = tcp_keepidle;
 321                 break;
 322
 323         dropit:
 324                 tcpstat.tcps_keepdrops++;
 325                 tp = tcp_drop(tp, 0); /* ETIMEDOUT); */
 326                 break;
 327         }
 328
 329         return (tp);
 330 }